joshua.decoder.ff.lm.AbstractLM Class Reference

Inheritance diagram for joshua.decoder.ff.lm.AbstractLM:

Collaboration diagram for joshua.decoder.ff.lm.AbstractLM:

List of all members.

Public Member Functions
	AbstractLM (int order)
final double	sentenceLogProbability (List< Integer > sentence, int order, int startIndex)
final double	ngramLogProbability (int[] ngram)
final double	ngramLogProbability (int[] ngram, int order)
final double	logProbOfBackoffState (List< Integer > ngram, int order, int qtyAdditionalBackoffWeight)
final double	logProbabilityOfBackoffState (int[] ngram, int order, int qtyAdditionalBackoffWeight)
int[]	leftEquivalentState (int[] originalState, int order, double[] cost)
int[]	rightEquivalentState (int[] originalState, int order)
Protected Member Functions
abstract double	ngramLogProbability_helper (int[] ngram, int order)
abstract double	logProbabilityOfBackoffState_helper (int[] ngram, int order, int qtyAdditionalBackoffWeight)

---

Detailed Description

This class implements NGramLanguageModel by creating wrappers around the necessary functions to capture common errors. Most methods are declared final, in an attempt to limit what subclasses may be defined.

Author:

Zhifei Li, zhife.nosp@m.i.wo.nosp@m.rk@gm.nosp@m.ail..nosp@m.com

Version:

$LastChangedDate$

---

Constructor & Destructor Documentation

joshua.decoder.ff.lm.AbstractLM.AbstractLM

(

int

order

)

---

Member Function Documentation

int [] joshua.decoder.ff.lm.AbstractLM.leftEquivalentState	(	int[]	originalState,
		int	order,
		double[]	cost
	)

Reimplemented from joshua.decoder.ff.lm.DefaultNGramLanguageModel.

Reimplemented in joshua.decoder.ff.lm.buildin_lm.LMGrammarJAVA.

final double joshua.decoder.ff.lm.AbstractLM.logProbabilityOfBackoffState	(	int[]	ngram,
		int	order,
		int	qtyAdditionalBackoffWeight
	)

Will never be called, because BACKOFF_LEFT_LM_STATE_SYM_ID token will never exist. However, were it to be called, it should return a probability of 1 (logprob of 0).

Reimplemented from joshua.decoder.ff.lm.DefaultNGramLanguageModel.

Here is the call graph for this function:

Here is the caller graph for this function:

abstract double joshua.decoder.ff.lm.AbstractLM.logProbabilityOfBackoffState_helper	(	int[]	ngram,
		int	order,
		int	qtyAdditionalBackoffWeight
	)		[protected, pure virtual]

Implemented in joshua.decoder.ff.lm.bloomfilter_lm.BloomFilterLanguageModel, joshua.decoder.ff.lm.buildin_lm.LMGrammarJAVA, and joshua.decoder.ff.lm.berkeley_lm.LMGrammarBerkeley.

Here is the caller graph for this function:

final double joshua.decoder.ff.lm.AbstractLM.logProbOfBackoffState	(	List< Integer >	ngram,
		int	order,
		int	qtyAdditionalBackoffWeight
	)

Deprecated:

this function is much slower than the int[] version

Reimplemented from joshua.decoder.ff.lm.DefaultNGramLanguageModel.

Here is the call graph for this function:

final double joshua.decoder.ff.lm.AbstractLM.ngramLogProbability

(

int[]

ngram

)

Reimplemented from joshua.decoder.ff.lm.DefaultNGramLanguageModel.

Here is the caller graph for this function:

final double joshua.decoder.ff.lm.AbstractLM.ngramLogProbability	(	int[]	ngram,
		int	order
	)		[virtual]

Implements joshua.decoder.ff.lm.DefaultNGramLanguageModel.

Here is the call graph for this function:

abstract double joshua.decoder.ff.lm.AbstractLM.ngramLogProbability_helper	(	int[]	ngram,
		int	order
	)		[protected, pure virtual]

Implemented in joshua.decoder.ff.lm.bloomfilter_lm.BloomFilterLanguageModel, joshua.decoder.ff.lm.buildin_lm.LMGrammarJAVA, and joshua.decoder.ff.lm.berkeley_lm.LMGrammarBerkeley.

Here is the caller graph for this function:

int [] joshua.decoder.ff.lm.AbstractLM.rightEquivalentState	(	int[]	originalState,
		int	order
	)

Reimplemented from joshua.decoder.ff.lm.DefaultNGramLanguageModel.

Reimplemented in joshua.decoder.ff.lm.buildin_lm.LMGrammarJAVA.

final double joshua.decoder.ff.lm.AbstractLM.sentenceLogProbability	(	List< Integer >	sentence,
		int	order,
		int	startIndex
	)

Parameters:

sentence	the sentence to be scored
order	the order of N-grams for the LM
startIndex	the index of first event-word we want to get its probability; if we want to get the prob for the whole sentence, then startIndex should be 1

Returns:

the LogP of the whole sentence

Reimplemented from joshua.decoder.ff.lm.DefaultNGramLanguageModel.